Abstract

This study investigates the influence of linguistic experience on the perception of phonation and the acoustic properties that correlate with this perception. Listeners from Gujarati (contrasts breathy vs. modal vowels), Spanish (no breathiness) and English (allophonic breathiness) participated in a free-sort task using stimuli from various languages/dialects that differ in their phonation production, a similarity-rating task using stimuli from Mazatec, a language that contrasts phonation, and a similarity-rating task with pathologically-disordered stimuli.

It was hypothesized that (1) due to the phonation contrast in Gujarati, Gujaratis would perform more consistently than English and Spanish listeners, (2) due to the allophonic breathiness in English, English listeners would be more consistent than Spanish listeners, (3) Gujarati and English listeners' judgments would correlate with the measure H1-H2 (amplitude of the first harmonic minus the second harmonic) because this measure correlates with their production of breathiness, and (4) Spanish listeners, in the absence of any guidance from their language, would rely on the largest auditory difference between the stimuli.

The results of the free-sort task were analyzed by determining per-pair consistency, correlations between acoustic measures and listeners' judgments, and percent of pairs sorted correctly. The results of the similarity rating tasks were analyzed using multi-dimensional scaling.

Results showed that Gujaratis did better at distinguishing breathy from modal vowels and were more consistent than either English or Spanish listeners. Despite their allophonic breathiness, English listeners did no better at distinguishing phonations than Spanish listeners did, nor were they more consistent. Gujaratis relied solely on the measure H1-H2, which is associated with the production of phonation in Gujarati. English listeners relied weakly on the measure H1-H2, which is associated with the production of phonation in English. English listeners also used cepstral peak prominence in the three similarity-rating tasks. Spanish listeners relied on H1-H2 and H1-A1 (amplitude of first formant peak).

In the similarity-rating task using pathologically-disordered stimuli, Gujaratis behaved more consistently than other listeners, who disagreed on the relevant importance of the measures that they used. All three listener types treated the pathologically-disordered stimuli in a similar way to the Mazatec stimuli.